CN115457476A - Continuous fiber 3D printing process monitoring method based on artificial intelligence image recognition - Google Patents

Abstract

A continuous fiber 3D printing process detection method based on artificial intelligence image recognition is characterized in that a process monitoring technology based on computer vision and mode recognition is selected to monitor the printing process of a continuous fiber reinforced composite material, a camera is used for collecting images of a continuous fiber printing piece, and a neural network classification model is combined to realize an image recognition classification function so as to realize automatic detection of printing defects in the printing process; according to the invention, on one hand, the defect can be automatically monitored in the continuous fiber 3D printing process, on the other hand, a foundation can be laid for subsequent real-time printing control or repair, so that the intelligent controllability of the printing quality is realized, the qualification rate of continuous fiber 3D printing products is improved, the material waste is reduced, and the manufacturing time is shortened.

Description

Continuous fiber 3D printing process monitoring method based on artificial intelligence image recognition

Technical Field

The invention belongs to the technical field of 3D printing of composite materials, and particularly relates to a method for detecting a 3D printing process of continuous fibers based on artificial intelligence image recognition.

Background

The 3D printing technology is originally called as a rapid prototyping technology, also called as an additive manufacturing technology, adopts a manufacturing process of manufacturing an entity from bottom to top by adopting a material layer-by-layer accumulation method, is different from the traditional material reduction manufacturing processes such as cutting processing and the like, and obviously reduces the requirements on the production of a mould and the limitation on the design complexity; the mechanical property of the composite material structure reinforced by the fibers is obviously improved, and the composite material structure has the advantages of high strength, high modulus, small specific gravity, corrosion resistance, good thermal stability and the like, and can be widely applied to multiple fields. With the development of the continuous fiber reinforced 3D printing technology, the accompanying quality problem is more and more prominent, and especially in the fields of rehabilitation, aerospace and the like, higher requirements are placed on the reliability and stability of the composite material 3D printing product.

The combination bonding of fiber and base member has more crucial influence to the quality and the performance of printing in the continuous fibers reinforcing 3D printing process, and a series of printing defect problems such as fibre route skew, fibre fracture, fibre are extracted and the base member fracture easily take place in the continuous fibers 3D printing process to cause combined material 3D to print the quality and the performance of piece and descend. Therefore, in order to improve the printing precision and ensure the comprehensive performance of the printed product, the process monitoring technology for the 3D printing of the composite material is very important.

Most of the printing defects detected in the existing research are related to the material extrusion amount (Zeqing Jin, automous in-situ correction of fused deposition modeling printers using computer vision and apparatus, manufacturing Letters, volume 22, 2019.); the method comprises the following steps of internationally and temporarily detecting and researching special fiber-matrix bonding defects in the 3D printing process of continuous fibers, and temporarily and continuously monitoring the additive manufacturing process by a multi-camera, wherein most of the objects to be researched and monitored are pictures of each layer in the 3D printing process, (Heke. 3D printing process fault diagnosis method and device are CN1099671B.2020-07-10; wangdi. A3D printing process monitoring method and device based on real-time camera shooting, CN106925784A.2017-07-07.); the conventional single camera follow-up printing process detection (William Jordan Wright, in-situ optimization of thermal composite Additive Manufacturing view de real and computer vision, additive Manufacturing, volume 58, 2022.), the information collection of the screen near the head is incomplete.

In summary, the lack of a process monitoring and defect identification method for a continuous fiber 3D printing technology leads to a composite material printed product often having quality problems such as fiber deviation, and limits the wide application of the composite material printed product in the industry.

Disclosure of Invention

In order to overcome the defects of the prior art, the invention aims to provide a continuous fiber 3D printing process detection method based on artificial intelligence image recognition, which can realize automatic monitoring of defects in the continuous fiber 3D printing process on one hand, and lay a foundation for subsequent real-time printing control or repair on the other hand, thereby realizing intelligent controllability of printing quality, improving the qualification rate of continuous fiber 3D printing products, reducing material waste and shortening the manufacturing time.

In order to achieve the purpose, the invention adopts the technical scheme that:

a continuous fiber 3D printing process detection method based on artificial intelligence image recognition is characterized in that a process monitoring technology based on computer vision and mode recognition is selected to monitor a continuous fiber reinforced composite material printing process, a camera is used for collecting images of continuous fiber printed products, and a neural network classification model is combined to realize an image recognition classification function, so that the automatic detection of printing defects in the printing process is realized.

The continuous fiber reinforced composite material comprises continuous fibers serving as a reinforcing material and resin serving as a base material, wherein the continuous fibers comprise carbon fibers, glass fibers, aramid fibers, flax fibers and the like; the resin includes polylactic acid (PLA), TPU, ABS, PEEK, PPS, etc.

The computer vision and pattern recognition uses double cameras to collect continuous fiber reinforced 3D printing process real-time images in horizontal and vertical directions at the spray head in a follow-up manner, and inputs the collected image data into an artificial intelligence algorithm which completes training in the early stage, so that real-time artificial intelligence image recognition is realized, namely, defects in the printing process are judged; the neural network classifier realizes output of defect monitoring in a 3D printing process for continuous fibers.

The artificial intelligence algorithm comprises a two-classification neural network model, a multi-classification convolutional neural network, a K value approximation algorithm and the like.

A continuous fiber 3D printing process detection method based on artificial intelligence image recognition comprises the following steps:

1) Determining a printing structure for acquiring a data set and an image acquisition mode corresponding to the expected judgment of the printing defects according to the continuous fiber printing platform structure and the defects expected to be monitored in the target printing process;

2) Designing and building a double-camera follow-up FDM continuous fiber 3D printing platform to realize the acquisition of image data in the printing process in the horizontal and vertical directions;

3) Importing the three-dimensional model file of the printing structure used for acquiring the data set in the step 1) into a 3D printer, and repeating the printing process to acquire image data;

4) The method comprises the steps of collecting images near a sprayer in the early stage, classifying image data according to specific situations of defects, and constructing a data set;

5) Training the initial neural network model by using the image data set in the step 4) to improve the precision and realize automatic diagnosis aiming at the printing defects;

6) And outputting a monitoring result aiming at the defects in the 3D printing process of the continuous fibers by the trained neural network model.

Compared with the prior art, the invention has the following beneficial effects:

(1) The invention combines the artificial intelligence technology with the composite material 3D printing technology, designs the composite material 3D printing process monitoring method based on the artificial intelligence image recognition technology, and can be used for judging the printing defects (such as whether the path of the continuous fiber of the reinforcement body is arranged in the middle of the base material) in the printing process, thereby improving the quality problem of the composite material printing piece caused by the printing defect problem.

(2) The invention provides innovative double-camera follow-up continuous fiber reinforced 3D printing process monitoring, at present, the defect detection of continuous fiber reinforced 3D printing is temporarily omitted internationally, the printing defects detected in the existing research are mostly related to material extrusion amount, the detection research of special fiber-matrix bonding defects in the 3D printing process of continuous fibers is temporarily omitted internationally, and the monitoring research of the material increasing and manufacturing process of the continuous fibers with follow-up multiple cameras is temporarily omitted at present; most of the objects of the existing research monitoring are pictures of each layer in the 3D printing process, but not the detailed image information near the spray head in the printing process, and the camera follow-up of the invention can solve the problem that the printing defect monitoring is not timely enough; the invention is different from the following printing process detection of a common single camera, simultaneously collects pictures in two directions, and can effectively improve the defect that complete information collection of pictures near a spray head cannot be carried out.

Drawings

FIG. 1 is a flow chart of monitoring a camera follow-up continuous fiber 3D printing process according to an embodiment of the present invention.

Fig. 2 is a schematic diagram illustrating whether a fiber path is shifted in a continuous fiber 3D printing process according to an embodiment of the present invention.

Fig. 3 is a schematic diagram of the neural network model for determining whether the fiber path is deviated according to the embodiment of the present invention.

Fig. 4 is a schematic diagram of a training effect of a neural network model according to an embodiment of the present invention.

Detailed Description

The invention is described in detail below with reference to the figures and examples.

Referring to fig. 1, a method for detecting a continuous fiber 3D printing process based on artificial intelligence image recognition includes the following steps:

1) Determining a printing structure for acquiring a data set and an image acquisition mode corresponding to the printing defect expected to be judged according to the continuous fiber printing platform structure and the defect expected to be monitored in the target printing process; in this embodiment, whether the fiber path is located in the middle of the matrix PLA and whether fiber-matrix debonding is performed are selected as examples;

2) In order to realize the monitoring of the composite material 3D printing process based on vision, a double-camera follow-up FDM continuous fiber 3D printing platform is designed and set up, and the collection of printing live pictures at the spray heads in the horizontal direction and the vertical direction is realized;

3) Importing the three-dimensional model file of the printing structure used for acquiring the data set in the step 1) into a 3D printer, and repeating the printing process to acquire image data;

in order to realize real-time monitoring of fiber path deviation in the FDM continuous fiber 3D printing process, firstly, a camera-following continuous fiber reinforced 3D printing platform needs to be built, and two cameras and a macro lens are matched to realize near-distance printed piece surface condition image information acquisition near a nozzle; carrying out structural design and modeling on the camera support by using three-dimensional modeling software, and manufacturing the camera support through a 3D printer; completing the structural design of a printing piece required by data set shooting and generating a 3D printing code;

the dual-camera follow-up design can realize the image acquisition of the positions near the spray head along the X and Y directions, and the neural network image classifier with artificial intelligence image recognition capability is obtained due to the need of training, the image information of the images in the continuous fiber printing process is required to be acquired as training data in the early stage, and then the neural network model is trained through machine learning;

4) Acquiring images near a sprayer in the early stage, manually classifying image data according to the specific condition of whether a fiber path deviates, and constructing a data set;

5) Training the initial neural network model by using the image data set in the step 4) to improve the precision and realize the printing defect diagnosis aiming at whether the fiber path deviates;

in the embodiment, the image data of the continuous fiber reinforced 3D printing process is acquired by adopting Action2 through dual-camera follow-up, and the initial video data is preprocessed; manually judging and classifying the collected images according to whether the fiber paths in the image number deviate or not, and referring to fig. 2, dividing the collected images into two categories of "GOOD" (the fiber paths do not deviate) and "BAD" (the fiber paths deviate); then randomly dividing the marked pictures into a training set and a testing set according to the proportion of 9; training an image classification initial neural network model by using a training set, and verifying the accuracy of the model in a test set, wherein 600 photos are intensively used in the training set;

6) The trained neural network model outputs a monitoring result aiming at whether the fiber path deviates in the continuous fiber 3D printing process, and the possibility is provided for realizing real-time condition monitoring in the subsequent composite material 3D printing process or subsequently carrying out a defect repairing function.

Referring to fig. 3, in the embodiment, an 18-layer ResNet is used as an initial neural network model for training, and through machine learning, the ResNet18 model learns feature points in various photos, so that the unmarked photos collected by the following camera are subjected to feature picking, and finally, the classification that the states of the printed matters are "GOOD" or "BAD" in the continuous fiber 3D printing process is realized. As shown in fig. 4, the initial neural network model is trained by using the data set classified according to the fiber path deviation condition, and in the process of completely traversing 10 data sets, the loss function is in a descending trend, the accuracy is in an ascending trend, and the effectiveness of the image recognition model training for whether the fiber path is deviated or not is reflected; after the training is finished, the average classification accuracy of 92% is achieved, and the trained neural network model is proved to have a good classification effect aiming at the characteristics.

According to the invention, by combining an artificial intelligence image recognition theory and a composite material 3D printing process, the printing defect recognition and judgment in the continuous fiber reinforced 3D printing process based on vision are realized, a foundation is laid for real-time control or repair aiming at the printing defect problem in the later period, the possibility is provided for the detection of the continuous fiber 3D printing process in an unmanned environment, and the method has potential application value in the fields of high-precision and high-performance printing requirements or unsupervised printing in aerospace, biomedicine and the like.

Claims (5)

1. A continuous fiber 3D printing process detection method based on artificial intelligence image recognition is characterized in that: the continuous fiber reinforced composite printing process is monitored by selecting a process monitoring technology based on computer vision and pattern recognition, images of continuous fiber printed products are collected through a camera, and the function of image recognition and classification is realized by combining a neural network classification model, so that the automatic detection of printing defects in the printing process is realized.

2. The method of claim 1, wherein: the continuous fiber reinforced composite material comprises continuous fibers serving as a reinforcing material and resin serving as a base material, wherein the continuous fibers comprise carbon fibers, glass fibers, aramid fibers and flax fibers; the resin includes polylactic acid (PLA), TPU, ABS, PEEK, PPS.

3. The method of claim 1, wherein: the computer vision and pattern recognition uses double cameras to collect continuous fiber reinforced 3D printing process real-time images in horizontal and vertical directions at the spray head in a follow-up manner, and inputs the collected image data into an artificial intelligence algorithm which completes training in the early stage, so that real-time artificial intelligence image recognition is realized, namely, defects in the printing process are judged; the neural network classifier realizes output of defect monitoring in a 3D printing process for continuous fibers.

4. The method of claim 3, wherein: the artificial intelligence algorithm comprises a two-classification neural network model, a multi-classification convolution neural network and a K value approach algorithm.

5. A method according to claims 1-4, characterized by the steps of:

1) Determining a printing piece structure for acquiring a data set and an image acquisition mode corresponding to the expected judgment of the printing defects according to the structure of the continuous fiber printing platform and the expected monitored defects in the target printing process;

2) Designing and building a double-camera follow-up FDM continuous fiber 3D printing platform to realize the acquisition of image data in the printing process in the horizontal and vertical directions;

3) Importing the three-dimensional model file of the printing structure used for acquiring the data set in the step 1) into a 3D printer, and repeating the printing process to acquire image data;

4) The method comprises the steps of collecting images near a sprayer in the early stage, classifying image data according to specific situations of defects, and constructing a data set;

5) Training the initial neural network model by using the image data set in the step 4) to improve the precision and realize automatic diagnosis aiming at the printing defects;

6) And outputting a monitoring result aiming at the defects in the 3D printing process of the continuous fibers by the trained neural network model.

Images